core: Bump the chunk size for ptr_table uses in types, tags, functions tables

On a:

  $ grep "model name" /proc/cpuinfo
  model name	: Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz
  model name	: Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz
  model name	: Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz
  model name	: Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz
  $

Before, using 256 as the chunk size:

  $ perf stat -r5 pahole -j --btf_encode vmlinux

   Performance counter stats for 'pahole -j --btf_encode vmlinux' (5 runs):

         8,336.57 msec task-clock:u        # 2.649 CPUs utilized   ( +-  0.19% )
                0      context-switches:u  # 0.000 /sec
                0      cpu-migrations:u    # 0.000 /sec
           69,028      page-faults:u       # 8.260 K/sec           ( +-  0.03% )
   28,799,380,143      cycles:u            # 3.446 GHz             ( +-  0.06% )
   66,068,272,802      instructions:u      # 2.29  insn per cycle  ( +-  0.00% )
   15,801,729,716      branches:u          # 1.891 G/sec           ( +-  0.00% )
      134,370,099      branch-misses:u     # 0.85% of all branches ( +-  0.07% )

          3.14696 +- 0.00527 seconds time elapsed  ( +-  0.17% )

  $

After bumping it to 1024:

  $ perf stat -r5 pahole -j --btf_encode vmlinux

   Performance counter stats for 'pahole -j --btf_encode vmlinux' (5 runs):

         8,255.93 msec task-clock:u        # 2.635 CPUs utilized   ( +-  0.03% )
                0      context-switches:u  # 0.000 /sec
                0      cpu-migrations:u    # 0.000 /sec
           68,597      page-faults:u       # 8.312 K/sec           ( +-  0.04% )
   28,504,209,806      cycles:u            # 3.454 GHz             ( +-  0.03% )
   66,067,020,098      instructions:u      # 2.32  insn per cycle  ( +-  0.00% )
   15,802,624,183      branches:u          # 1.915 G/sec           ( +-  0.00% )
      133,542,603      branch-misses:u     # 0.85% of all branches ( +-  0.13% )

          3.13324 +- 0.00205 seconds time elapsed  ( +-  0.07% )

  $

And 2048:

  $ perf stat -r10 pahole -j --btf_encode vmlinux

   Performance counter stats for 'pahole -j --btf_encode vmlinux' (10 runs):

         8,237.37 msec task-clock:u        # 2.635 CPUs utilized   ( +-  0.02% )
                0      context-switches:u  # 0.000 /sec
                0      cpu-migrations:u    # 0.000 /sec
           68,643      page-faults:u       # 8.331 K/sec           ( +-  0.06% )
   28,447,701,874      cycles:u            # 3.453 GHz             ( +-  0.02% )
   66,077,728,879      instructions:u      # 2.32  insn per cycle  ( +-  0.00% )
   15,806,113,927      branches:u          # 1.918 G/sec           ( +-  0.00% )
      132,811,965      branch-misses:u     # 0.84% of all branches ( +-  0.11% )

         3.125675 +- 0.000905 seconds time elapsed  ( +-  0.03% )

  $

Value chosen using:

  $ pahole --ptr_table_stats --btf_encode vmlinux

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
1 file changed